PyGWalker 0.6 introduces the new public computation model and the reusable Walker API while keeping the 0.5 user-facing computation flags available for compatibility.
Use computation to choose where queries run:
pyg.walk(df, computation="auto")
pyg.walk(df, computation="browser")
pyg.walk(df, computation="kernel")
pyg.walk(df, computation="cloud")For code that renders the same data in more than one environment, prefer a reusable Walker:
w = pyg.Walker(df, computation="kernel")
w.show()
w.to_html()
w.to_streamlit()kernel_computation, cloud_computation, and use_kernel_calc are deprecated in 0.6 but remain supported. They emit DeprecationWarning and are scheduled for removal in PyGWalker 0.7.0. Do not mix an explicit non-auto computation value with enabled legacy computation flags.
The anywidget path is the preferred and default Jupyter rendering path. Jupyter and JupyterWidget remain accepted as legacy aliases in 0.6, but they warn and resolve to the anywidget transport. These aliases are scheduled for removal in PyGWalker 0.7.0.
Static HTML export is browser-only. Exporting static HTML from computation="kernel" or computation="cloud" raises an error instead of producing output that cannot execute live kernel or cloud computation.
The default privacy mode is update-only. Event telemetry is sent only when privacy is set to events, and the frontend Segment tracker opens only for that events setting.
The 0.6 release line is validated against Python 3.10, 3.11, 3.12, and 3.13. The frontend and publish workflows build with Node.js 22.x, and CI runs the frontend build plus Playwright smoke tests before packaging.